Nested species distribution models (SDMs) can account for the multi-scale nature of species' niches yet typically combine only two spatial scales. Here, we present a novel meta-model approach for nesting more than two scales, which is particularly relevant for mobile species that select habitat across different spatial scales. We demonstrate the approach by modelling two farmland bird species in Germany, the common starling (Sturnus vulgaris) and the northern lapwing (Vanellus vanellus). First, we trained three single-scale SDM ensembles: a European-extent climate ensemble at 50×50 km, a German-extent landscape context ensemble at 1×1 km, and a German-extent fine-scale habitat ensemble at 200×200 m. Using stacked generalization, we combined the single-scale predictions into a multi-scale meta-model at 200×200 m resolution using ridge regression and quantified each scale's relative importance as the reduction in explained deviance when excluding it from the meta-model. Although all single-scale SDMs performed well, the multi-scale SDM further improved predictive performance. Across both species, fine-scale habitat was the most relevant scale, followed by landscape context, with broad-scale climate being the least important. The meta-model did not introduce novel spatial predictions, but rather combined and reinforced patterns identified at individual scales. Our approach offers a flexible and reproducible framework for integrating multiple spatial scales into SDMs and quantifying their relative importance, advancing both the predictive accuracy and ecological interpretability of SDMs.